Thanks for the links. Here's my thoughts, take them or leave them: First, it's a big leap to go from the Harvard article's "here are theoretical issues with this test, not knowing any specifics of how it works" to "this doesn't work." There appear to be two components to the test, a genetic admixture, and predicted phenotypic characteristics that manifest with specific allele groupings, within a certain threshold of accuracy. Genes are not all equally heritable, there are dominant characteristics, so it can't be said that if someone is 50% one ethnicity and 50% another that you can't make any predictions about their appearance. This doesn't represent a rigid genetic definition of "race" I agree, but if it works for its intended purpose (matching phenotypic characteristics we generally associate with race to a genetic makeup) it opens the door to doing a blood test, finding someone's peak intelligence (assuming intelligence is heritable and measurable) according to some metric, and, possibly, associating that with a heritage/ethnicity/declared race, again within a certain accuracy threshold. (On the other hand, it may not extend all the way to heritable intelligence, or, the blood test might not even work for its intended purpose.) I also agree this would never mean that someone could say to an individual "you are a member of this race, therefore your intelligence peaks at IQ X" but you could use it to make predictions about an ethnic group in the aggregate.

I tried to qualify my statements here very carefully. I'm not claiming a strict genetic definition of race. I'm also not defending this specific blood test, I only know about it from that Wired article. But the theoretical possibility of what I have mentioned seems to exist, if that test or one like it actually works. Or am I mistaken?